mirror of
https://github.com/Routstr/routstr-core.git
synced 2026-10-05 20:28:23 +00:00
redshift
6c91eac02c
The cache discount still did not apply after 3c813da because the EHBP
finalization path never handed the serving ``Model`` to ``calculate_cost``:
it passed only the model *string* (``{"model": pricing_model_id, ...}``),
so ``_get_pricing_rates`` fell into its "settling without routed model
identity" branch and re-derived pricing through the global alias map.
That lookup resolves the string to the *best-ranked* candidate for the
id, not the serving one. Tinfoil's catalog id (``deepseek-v4-1-flash``)
is also a bare cross-provider alias, and the cheaper cross-provider
candidate carries ``input_cache_read=0``. ``_get_pricing_rates`` treats a
zero cache rate as "missing" and falls back to the full input price, so
every request billed at the undiscounted rate (and at the cross-provider
price, ~28% low): production showed ``Applied model-specific pricing:
input=494.55, cache_read=494.55`` with ``source=configured`` even though
the enclave reported ``cached_prompt_tokens=12800``.
The sibling path is correct: ``BaseUpstreamProvider.get_x_cashu_cost``
already passes ``model_obj`` to ``calculate_cost``. The EHBP path is the
only caller that omitted it.
Fix: thread the pricing model object through ``_compute_ehbp_actual_cost``
— the routed ``model_obj`` normally, or the resolved served model on a
genuine mismatch — and pass it to ``calculate_cost``. The requested model
string stays in ``response_data["model"]`` for logging. 3c813da's
namespace resolution is still what decides *which* object to bill; this
commit makes that object actually reach pricing.
Verified numerically in the new regression test: 12800 cached tokens at
the Tinfoil cache rate (~112 msat/1k) bill ~1434 msats instead of ~8800
msats at the full input rate.
Tests: 2 new in tests/unit/test_tinfoil_integration.py; full suite 2001
passed, 14 skipped. ruff check + mypy clean.
FastAPI Async Unit Tests
This directory contains async unit tests for the Routstr proxy FastAPI application.
Installation
First, ensure you have the development dependencies installed:
uv pip install -e ".[dev]"
Running Tests
To run all tests:
pytest
To run tests with coverage:
pytest --cov=routstr --cov-report=html
To run specific test files:
pytest tests/test_main.py
pytest tests/test_models.py
pytest tests/test_proxy.py
To run only async tests:
pytest -m asyncio
Test Structure
conftest.py- Pytest fixtures and configurationtest_main.py- Tests for main app endpointstest_account.py- Tests for wallet/account management endpointstest_proxy.py- Tests for the proxy functionality with mocked upstreamtest_models.py- Tests for model pricing and data structures
Key Fixtures
async_client- Async HTTP client for testing FastAPI endpointstest_session- In-memory SQLite database session for teststest_api_key- Pre-configured API key with balanceapi_key_with_balance- API key with sufficient balance for proxy tests
Environment Variables
The tests automatically set up required environment variables in conftest.py. No manual configuration needed.
Writing New Tests
- Use
@pytest.mark.asynciofor async tests - Use the provided fixtures for database and client access
- Mock external dependencies (like upstream API calls)
- Test both success and error cases
- Verify database state changes when applicable